PDF to Excel
Recover tables from a PDF into a real spreadsheet.
About PDF to Excel
Turn the tables in a PDF back into rows and columns you can sort, total and edit, exported as a genuine .xlsx workbook with one sheet per page. Columns are found from where the text actually sits on the page, and nothing is uploaded to do it.
The table you need is right there on the page, and selecting it gives you a wall of run-together text. That is not a failure of the copy command — it is what a PDF is. The file records which glyph sits at which coordinate; the rows and columns were a property of whatever produced it, and were discarded on the way out.
So the grid has to be worked out again from the only evidence left, which is where the text sits. That is inference rather than conversion, and it is why this tool shows you the result before it offers you a file.
- Processed on your deviceEverything you open or type here is handled by this page and never uploaded anywhere.
- No queue, no limitsNothing waits on a server, so the only limit is your own hardware.
- No account neededNo sign-up and no email address required.
How PDF to Excel works in your browser
PDF.js reports every run of text on a page with its position, size and width. Runs sharing a baseline are assembled into a row, and a horizontal gap wider than a word space splits that row into cells.
Columns are then found across the whole page rather than row by row, because a single row cannot tell you anything: two figures with a space between them could be two columns or one sentence. Every cell’s left edge is treated as a vote, edges within a few points of each other are counted as the same position, and a position only becomes a column once enough separate rows agree on it. An indent that occurs once is an indent; one that recurs forty times down the page is a column.
Figures are turned into real numbers where that is unambiguous — thousands separators removed, a leading currency symbol stripped, accounting parentheses read as a negative. A percentage keeps its sign rather than quietly becoming a bare number, and anything that does not parse cleanly is left as text. Guessing wrong about a figure corrupts it silently, which is worse than leaving a column for you to format.
Three things it cannot do, all for the same reason. Merged cells have one position, so a merged value lands in the first column it starts in. A cell whose text wrapped arrives as a second row, because visually that is what it is. And a scanned page has no text at all, so there is nothing to measure — the tool says so rather than producing an empty sheet.
When to use it
Getting a bank or credit-card statement into a spreadsheet so the year can be totalled, when the bank only offers PDF. Recovering the figures from a report whose source workbook nobody kept. Pulling a price list out of a supplier’s catalogue rather than retyping two hundred rows. Extracting a results table from a published paper to check the arithmetic.
Frequently asked questions
How does it know where the columns are?
From geometry. A PDF stores no table structure at all, so the left edge of every cell on the page is measured and the positions that recur down the page are taken as the columns. A position that appears on one row only is treated as an indent rather than a column.
Will it work on a scanned document?
No. A scan is a photograph of a table, with no text to position, and there is no OCR here. The tool will tell you the page has no text layer rather than hand back an empty sheet.
What does it get wrong?
Merged cells arrive in the first column they start in, and a cell whose text wrapped onto a second line arrives as a second row. Check the preview before you download — it shows exactly what the workbook will contain.